56812: kvserver: replace multiTestContext with TestCluster in client_raft_test r=lunevalex a=lunevalex Makes progress on #8299 Fixes #40351 Fixes #57560 Fixes #57537 multiTestContext is a legacy construct that is deprecated in favor of running tests via TestCluster. This is one PR out of many to remove the usage of multiTestContext in the client_raft test cases. This does not remove all the uses of mtc, just the simple ones. Leaving the more complex uses cases for a later PR. With this switch we can also clean up some TestingKnobs and TestServer interfaces. - DisablePeriodicGossips flag is removed, it does not work with TestCluster and is no longer used - DontPreventUseOfOldLeaseOnStart flag is removed, it did not work consistently in TestCluster. This flag tries to leave the Lease on the same node after a restart, but CRDB makes no such guarantees in the real world and artificially testing it does not prove anything. The affected tests were re-worked to not rely on this condition and can deal with a lease holder moving on a restart. - GetRaftLeader is ported from multiTestContext to TestCluster Release note: None 58265: sql: fix substring(byte[]) to treat input as raw bytes without escaping r=solongordon a=rafiss fixes #57367 Release note (bug fix): The substring function on byte arrays would treat its input as unicode code points, which would cause the wrong bytes to be returned. Now it only operates on the raw bytes. Release note (bug fix): The substring(byte[]) functions were not able to interpret bytes that had the `\` character since it was treating it as the beginning of an escape sequence. This is now fixed. 58902: tracing,testutils: detect span leaks r=irfansharif a=irfansharif With always-on tracing, we're maintaining in-memory registry of active spans (#58490). Spans are added and removed from this registry when they're Start()-ed and Finish()-ed. Spans that not explicitly finished (typically using `defer sp.Finish()`) are now a resource-leak, as they take up room in the registry. We'll want to find instances of this leak as soon as they crop up. To that end we add a check in our TestCluster.Stop codepaths that asserts against the registry being empty. This should give us wide coverage given it's usage throughout. We expect this change to capture the cases described in #58721. --- This check is currently failing, given we're actually leaking spans in `txnState.resetForNewSQLTxn`. Fixing that doesn't look simple, it ties into questions around draining SQL and rolling back open txns. Release note: None 59049: sql: implement alter sequence/view owner r=arulajmani a=RichardJCai Resolves #57965 Release note (sql change): Add support for ALTER VIEW/SEQUENCE OWNER TO commands. 59055: colexec: fix external aggregator fallback and bool agg functions reset r=yuzefovich a=yuzefovich Previously, `reset` method of the ordered aggregator would always set the flag to reset the internal batch to `true`. However, that batch is only allocated when `Next` is called at least once with a non-zero batch coming from the input, which is not the case when the fallback to the disk-backed strategy occurs in the external aggregator (there, we call `reset` before we use the operator every time). This would lead to a nil pointer exception, and it is now fixed. Our unit tests didn't catch it because we forgot to set the forced number of repartitions which is now also fixed. This also revealed a bug with resetting of `bool_and` and `bool_or` aggregates - we forgot to reset whether they have seen a non-null value or not. Fixes: #59043. Release note: None (no stable release with these bugs) Co-authored-by: Alex Lunev <alexl@cockroachlabs.com> Co-authored-by: Rafi Shamim <rafi@cockroachlabs.com> Co-authored-by: irfan sharif <irfanmahmoudsharif@gmail.com> Co-authored-by: richardjcai <caioftherichard@gmail.com> Co-authored-by: Yahor Yuzefovich <yahor@cockroachlabs.com>

kv: TestReplicateAfterTruncation is flaky #57537

Description

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions